Scholarlink

<p>Efficient gas-lift allocation in dual-string well completions remains a critical challenge in enhancing oil production from mature fields. Conventional trial-and-error approaches are often inefficient, resulting in suboptimal injection strategies and increased operational costs. This study presents an integrated data-driven framework that combines multiple linear regression (MLR), Random Forest (RF), and Bayesian optimization (BO) to optimize gas-lift injection allocation in dual-string completions. Field implementation results demonstrate a reduction in total injected-gas volume from 1.85 to 1.43 MMscf/d, representing a 22.7% decrease in gas consumption and an 11.3% increase in liquid production following optimization. The MLR model achieved <i>R</i><sup>2</sup> = 0.991 and RMSE = 0.891, while the RF model produced <i>R</i><sup>2</sup> = 0.837 and RMSE = 0.755, indicating that despite the higher <i>R</i><sup>2</sup> of MLR, RF offered lower prediction error and superior generalization to field data. The dataset covering 2014–2022 was processed via the WinGLUE® platform with CO<sub>2</sub> tracer diagnostics to ensure industrial validity. The proposed framework provides a robust and scalable approach for gas-lift optimization, enabling improved production efficiency and reduced operating expenditure in mature oil fields.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Using Statistical Methods to Optimize Gas-Lift Injection Allocation in Dual-String Well Completions

  • 期刊论文

  • 发表日期：

    2025年12月13日

查看全文
  • Nur Aina Hannani Kasranuddin,
  • Bashir Busahmin

摘要

Efficient gas-lift allocation in dual-string well completions remains a critical challenge in enhancing oil production from mature fields. Conventional trial-and-error approaches are often inefficient, resulting in suboptimal injection strategies and increased operational costs. This study presents an integrated data-driven framework that combines multiple linear regression (MLR), Random Forest (RF), and Bayesian optimization (BO) to optimize gas-lift injection allocation in dual-string completions. Field implementation results demonstrate a reduction in total injected-gas volume from 1.85 to 1.43 MMscf/d, representing a 22.7% decrease in gas consumption and an 11.3% increase in liquid production following optimization. The MLR model achieved R2 = 0.991 and RMSE = 0.891, while the RF model produced R2 = 0.837 and RMSE = 0.755, indicating that despite the higher R2 of MLR, RF offered lower prediction error and superior generalization to field data. The dataset covering 2014–2022 was processed via the WinGLUE® platform with CO2 tracer diagnostics to ensure industrial validity. The proposed framework provides a robust and scalable approach for gas-lift optimization, enabling improved production efficiency and reduced operating expenditure in mature oil fields.